Echo

[rwarshawski]

Anybody seen this?
 
http://en.wikipedia.org/wiki/Amazon_Echo
 
 
How easy would it be to hack?
Seems like a good extension for HAL.
 
bob

 

[randyth]

Here's a fellow who's discovered some interesting info: http://www.piettes.c...mazon-echo-api/
 
Similarly, here's a guy who is constantly polling the Echo's to-do list to act on voice "commands": https://www.youtube....h?v=r58ERvxT0qM
 
I'm hoping Amazon opens up the API, at least to the HA manufacturers like Universal Devices, so the different systems can reliably communicate.  The Echo has truly excellent voice recognition.
 
-Randy

 

[pvrfan]

http://www.geekculture.com/joyoftech/joyarchives/2069.html
 

 
Craig

 

[pete_c]

I ordered one mostly to tinker with.  The estimated date of delivery though is April for mine.  I hope I will get it sooner.
 
Here's an interesting youtube video which I posted on the Homeseer automation forum.
 
http://youtu.be/GijLoiVkmYI

 

[linuxha]

hehe, nice choice, the Samuel L. Jackson version wasn't as funny though.

I have no intention of getting one of these, the always listening part scares me (kind of like cell phones and laptops and tablets and things).

 

[ano]

I'm getting one.  And Siri on an iPhone can be set to listen all the time as well. It works pretty good.

 

[fcwilt]

I've got one. I like it. The things that it is programmed to respond to are somewhat limited. It's speech recognition is very good. It had no real problems with anyone in the family, anywhere in the room.
 
For example, it was rather nice to be able to ask about the weather as opposed to going online or firing up the TV.
 
I did use the Timer feature quite a bit and the Alarm feature a few times. The current limit of one Timer and one Alarm needs to be lifted.
 
The grand kids got a kick asking math questions. 
 
It works much better then I thought it would and I actually found myself using it most every day.

 

[ano]

I really like mine. It is very handy. Sure its version 1, but it is a very good start. Now, if it can control home automation and the TV, it would be great.
 
I'm very impressed with the voice recognition. I have mostly floors in my house which causes echos, and this thing even works 30 feet away. It has a variety of microphones and it pinpoints your location when you talk. I've also never had any false triggers.

 

[NeverDie]

I let my invitation expire, but your guy's enthusiasm for it got me to request another invite.
 
Does it recognize words better than the Kinect?
 
Our family doctor uses Dragon Dictate with a headset to transcribe her file notes, and watching over her schoulder I was impressed at just how accurate it had become.
 
Is there some breakthrough technology that's being applied to make voice recognition far more accurate now than it was formerly?

 

[wuench]

There are videos on youtube demonstrating the far field mic capabilities, people walking out of the room with background tv playing and no issues.  It is pretty impressive  I wish someone would just produce the far field mic tech in a way it can be used with other systems.  I let my invite expire as well.  I sent them a developer request but never heard back.
 
My original xbox 360 kinect was pretty good, the xbox one kinect is horrible at recognition.  I often have to yell at it half a dozen times to get it to turn on.  My Moto X works pretty well but you have to be within 6 inches for it to recognize/understand the command.  I have that controlling pretty much everything with CQC/Tasker.

 

[pete_c]

I am still waiting on my Echo order. 
 
That said meanwhile did purchase a few refurbished Kinect one's to play with.  I did also purchase a test mini pc that will run Wintel 8.1 for connectivity to the Kinects.  (so far only have 4 Kinects to play with).  The purpose of the mini pc will be only for use with the KInects and connect to Homeseer running in Linux.
 
The mini pc specs and box included a copy of WIndows 8.1 for some ~$90 USD.
 
Mini PC
CPU     Intel Baytrail Z3735F or Z3736F
CPU     2.16Ghz(Z3736F)
WIndows 8.1
 
Still waiting on that too; which now is becoming a bit of an annoyance.

 

[ano]

Is there some breakthrough technology that's being applied to make voice recognition far more accurate now than it was formerly?

Voice recognition can either be done locally or in the cloud. When you use voice recognition on a PC or even with Dragon, it is performed locally. The Echo performs it over the cloud, the same way Suri operates.  This let devices with relatively low processing power (like the Echo) perform a complex job of voice recognition. The negative is what you say is always transmitted to a third party where others can access it. But the positive is that cloud processing can be very sophisticated. 
 
The Echo is very rarely wrong on the voice recognition part. It keeps a log of this so you can see what it thinks it heard. The part that is even trickier is translating your English into a response.  Like my wife keeps giving the Echo compliments. It understands them perfectly, but it doesn't know how to respond to them.  

 

[ano]

I wish someone would just produce the far field mic tech in a way it can be used with other systems.  I let my invite expire as well.  I sent them a developer request but never heard back.

Voice tracker mics have been out there for years. I used this one with CQC maybe 10 years ago. https://www.acousticmagic.com/
It worked well, but the limitation was voice recognition on the PC. What Siri and Echo uses is cloud voice recognition. So a local device will need to send a voice segment to a service bureau for processing. Those exist as well, but usually for a fee.
 
But even with a mic, and a device to send it off to be processed, and a voice processing service bureau, you still need a way to take the reply, and do something with it, like "Turn on a kitchen light" and translate it into UPB device 102 on 100%.  Its not that it can't be done, just lots of moving pieces.

 

[Swancoat]

I'm not 100% sure Siri uses cloud recognition anymore. It definitely used to, and it definitely (obviously) uses cloud language translation to answer your query.

Just voice recognition/dictation seems like it's on-device now. Try to use the mic in they keyboard to dictate a text message. It used to just show a 'busy' animation as you spoke, waited until you were done - waited a few more seconds and then - boom. Filled in the text it thought it heard.

In iOS 8 though, it's basically populating word in real-time as you speak. That seems more like on-device speech recognition.

 

[Swancoat]

Uh, never mind. Just tried to dictate a short note with Airplane Mode on. No dice.

They sure made the cloud processing a lot faster I guess.